Policy-backed infrastructure for governed AI action
AI Workspace, Warrant and Warrant MCP address different layers of governed AI operation. These pages separate what each product does, what has been verified, and where its boundary sits.
Core platform
Operating layer
Commerce authorization
Warrant
Agent tool policy
Warrant MCP
- One System Architecture
- Shared Governance
- Protocol Native
- Portable Evidence
- Human Authorization
One thesis, two authorization products
Human policy becomes machine-enforceable structure, and deterministic code makes the runtime authorization decision.
Warrant applies deterministic authorization to AI purchasing proposals. Warrant MCP applies deterministic policy checks to supported AI-agent actions and tool calls. They share a design thesis, not a claimed shared runtime platform.
Verified · CL-48Portfolio
Choose the authorization boundary
Warrant
Commerce-specific authorization for AI purchasing agents.
Warrant compiles human-written spending policy into numbered clauses, requires human confirmation, and evaluates purchase proposals with deterministic ALLOW, ESCALATE or DENY decisions that cite the governing clause.
Verified · CL-40Warrant MCP
Deterministic policy enforcement for AI-agent actions and tool calls.
Warrant MCP compiles plain-English policy once into numbered clauses and closed structured rules. Deterministic code returns ALLOW or DENY for a supported action, and a Claude Code PreToolUse hook makes DENY a hard block.
Verified · CL-44Core platform
AI Workspace
Organizational knowledge, governed assignments, execution, verification and review.
Explore the platformPut the product family in front of your team
Every product shares one governance model, so evidence travels with the work.